From: Pyron Carol
Sent: Tuesday, February 02, 2010 12:08 AM
Subject: 1149.1 INITIALIZE Sub-Group Meeting Notes Jan 29 2010
Attachments: Rules_v0 11.rtf
Carl Barnhart (without microphone)
Sivikumar Jaya (do I have your name correct?)
CJ Clark sent notice that he was unable to attend due to a meeting conflict.
We continued the discussion of the rules. Just before the meeting, Carol sent out a marked up version of the Rules (version v0.10) and updated slides on "Some Persistence After TRST*". This was based on the previous week's discussion.
We completed the review pass through the Rules with changes to require TRST to not clear registers to allow some support for sequences of EXTEST - TRST* - EXTEST ... after a beginning Initialization Process. Potentially require extensive INIT processes between each set of EXTESTs a board test may consist of could be burdensome to board-level test times. Rules were changed to prevent reset of the INIT_DATA. Since the meeting Carl Barnhart has distributed additional views on this.
From a chip implementation point of view, there may not be any stable or well defined Ready-for-Test state. Any input could disturb this state as there may be multiple ways to "wake-up" or trigger actions on the chip. It would be very difficult or basically impossible on at least some chip architectures to only enable resets to work in this mode base don a clear need to establish priority of operation during "normal" power-up or boot sequences. Complexities of third-party IP can cause further issues. There was no clear path of feasibility within reasonable burdens to chip designers to convert Ken Parker's conceptual Read-to-Test mode to true and viable chip states.
We discussed further some of needs for the descriptive text. Noted that we should state that system resets are blocked in EXTEST or other intrusive instructions.
Ken Parker noted to the group that he plans to submit a paper to ITC as an individual (not a committee member) to describe in detail the difficulties caused over the years by the inherent "lobotomized" state of chips after a Test-Logic-Reset.
We reviewed the previous work on BSDL extension briefly and agreed the BSDL changes still looked to be small overall.
We spent the last part of the meeting with a general purpose discussion on the high-level requirements of the "side-file" to hold the INIT_DATA values.
Discussion and general (not-voted-on) consensus was:
- We should try to re-use an existing standard or a subset of the standard
- Side files should support more than a simple, dull binary or Hex string
- Side files will be man-made and man-read and they also be machine-made and machine read
- Some form of at least optional header fields or at least comment capability to aid in configuration management would be good
- Side files should optional support sub-fields or mnemonics for the binary string
- No need for "procedural" info
- Various possible sources we could "borrow" from included:
- TCL strings
- Forms demo'ed by CJ back in Dec
- more to come ...
We will discuss side file options more in the next meeting.
The updated rules are attached.
Please send me any changes or corrections.